memory-cli

Storing and retrieving contextual memories for AI agents from the command line involves using a CLI tool to save feedback, user data, and project notes, then querying them via fast full-text SQLite/FTS5 searches or local semantic embeddings.

Managing user feedback and project decisions for continuous agent improvement is best handled by storing these contextual entries in a structured SQLite database, allowing you to easily search and reference past interactions during future tasks.

Performing semantic search on stored memories using local embeddings requires an optional local embedding model integrated with your database, enabling you to retrieve contextually similar feedback or project notes beyond exact keyword matches.

Yes, storing agent sessions and memories requires a SQLite/FTS5 database to enable full-text search capabilities, along with an optional local embedding model if you want to perform semantic search queries.

Starting and ending scoped sessions to capture context for complex tasks is done directly through CLI session management commands, ensuring all feedback and project decisions are recorded within the correct task scope.

Refreshing indexes and rebuilding embeddings for a SQLite memory database is accomplished using built-in maintenance utilities, which apply database migrations and update search structures to keep memory retrieval optimized.